Power and Type-I Error in a Global Test of Differential Genetic Expression
Methods have been proposed to test for differential gene expression in microarray experiments. These include gene-specific tests and global-type tests, such as fitting mixture models to the distribution of test statistics or p-values or fitting an ANOVA model and testing for a gene-treatment interaction effect. This talk focuses on the latter for a particular microarray experimental design and discusses the effects of violations of ANOVA assumptions (e.g., correlated expressions, unequal variances, and nonnormal data) on type-I error and power. The residual bootstrap is considered as a way to compensate for effects of these violations. Finally, the distributions of gene-specific p-values are compared when computed using usual t-tests versus using effects estimated from the ANOVA model.
G. L. Gadbury et al., "Power and Type-I Error in a Global Test of Differential Genetic Expression," Proceedings of the 2005 Joint Statistical Meetings, American Statistical Association, Jan 2005.
Mathematics and Statistics
Keywords and Phrases
anova; microarray; p-valve; power
Article - Conference proceedings
© 2005 American Statistical Association, All rights reserved.
01 Jan 2005